Estimation of wooden cross-arm integrity using artificial neural networks and laser vibrometry

A significant problem faced by utility operators is the degradation and failure of wooden cross-arms on transmission line support structures. In this paper, a nondestructive, noncontact, reliable method is proposed, which can quickly and cost-effectively evaluate the structural integrity of these cross-arms. This method utilizes a helicopter-based laser vibrometer to measure vibrations induced in a cross-arm by the helicopter's rotors and engine. An artificial neural network (ANN) then uses these vibration spectra to estimate cross-arm breaking strength. The first type of ANN employed is the feed-forward artificial neural network (FFANN). After proper training, the FFANN can reliably discern healthy cross-arms from those that are in need of replacement based on vibration spectra. Next, a self-organizing map is applied to this same problem, and its advantages are discussed. Finally, a FFANN-based data compression scheme is presented for use as a preprocessor for the vibration spectra.